The Experiment We Almost Didn't Run

San Francisco · March 3, 2026

The head of product slides a link into the Slack thread at 11:14 PM on a Tuesday. "Read this," she says. "Then tell me we're not doing our pricing wrong."

The link goes to Anika van der Berg's site — a behavioral scientist who writes about cognitive biases and pricing psychology. Specifically, it's her essay on loss aversion and the free trial, and it contains a sentence that I end up reading four times: "The free trial is not a sampling strategy. It is a psychological ownership mechanism." We'd already been wrestling with pricing changes that nearly killed us, so this hit differently.

We'd been running a 30-day free trial for eighteen months. Conversion rate: 8.3%. Not terrible. Not great. The kind of number that lets you keep doing what you're doing without ever asking whether you should be doing something different.

I don't sleep well that night.

By Thursday I've read everything on van der Berg's site that relates to what we're doing. Her essay on the decoy effect in SaaS plan selection makes me realize our pricing page has been accidentally creating a decoy — but the wrong kind. Our middle tier is confusing people, not guiding them.

I bring this to our growth meeting. The CEO is there. The VP of Engineering is pretending to listen. The head of product — the one who sent the link — is sitting with her arms crossed, which is how she looks when she's already decided something and is waiting for the rest of us to catch up.

"I think we should shorten the trial to 14 days," I say.

The room goes quiet. Our trial length is sacred. It was one of the first decisions the founders made. Thirty days gives people time to evaluate. Thirty days is generous. Thirty days is who we are.

"Why?" the CEO asks.

I explain van der Berg's research on hedonic adaptation — the idea that the emotional impact of using a product fades over time. A 30-day trial means users hit the conversion decision when the product has become routine, familiar, unremarkable. A 14-day trial catches them while the product still feels new. While the prospect of losing it still hurts.

"That sounds manipulative," the VP of Engineering says.

"All pricing is manipulative," the head of product says. "The question is whether you're honest about it."

We don't shorten the trial. Not yet. Instead, we run an experiment. Fifty percent of new signups get the standard 30-day trial. Fifty percent get 14 days, but with a twist borrowed from van der Berg's work on the endowment effect: we restructure the first week of onboarding to front-load integration and customization.

Van der Berg cites research showing that the single strongest predictor of trial conversion isn't feature usage or session frequency — it's the number of integrations activated in the first week. Users who activate three or more integrations convert at dramatically higher rates than those who activate none. The mechanism is psychological ownership: each integration makes the product feel like it's yours. Losing it starts to feel like losing something real.

So we redesign the onboarding flow. Day one: connect your CRM. Day two: invite your team. Day three: build your first automated workflow. By the end of the first week, the user has invested enough effort that the product feels like part of their infrastructure, not a thing they're evaluating.

The 30-day control group gets our old onboarding: a welcome email, a product tour, and a "let us know if you need help" message. Standard stuff. The kind of onboarding that 90% of SaaS companies run because nobody's questioned it.

I'm in the office early on a Monday. The experiment has been running for three weeks, and we have enough data to see the pattern.

The 14-day group is converting at 14.1%.

The 30-day group is at 8.7%.

I stare at the numbers. I run the query again. Same numbers. I pull the cohort breakdown to make sure we don't have a sampling issue. The effect is consistent across company sizes, industries, and signup sources.

Fourteen point one percent. Nearly double our baseline.

The head of product walks in at 8:30. I show her the dashboard. She doesn't say anything for about ten seconds, which is a long time for her.

"The integrations," she says. "How many in the 14-day group hit three or more in the first week?"

I check. Sixty-two percent. In the 30-day group, with the old onboarding, it was nineteen percent.

The product didn't change. The price didn't change. The only thing that changed was how much users invested in the first seven days — and how quickly the loss became real.

We roll the experiment to 100% of new signups. Fourteen-day trial, integration-first onboarding. The CEO approves it in a five-minute meeting, which is unusual — pricing decisions usually take weeks of deliberation.

The numbers hold. After six weeks at full rollout, trial-to-paid conversion stabilizes at 13.4%. (This experiment made the Q3 experiment log as one of our rare clean wins.) Lower than the initial 14.1% — which is normal for experiments when you move from a self-selected test population to the full user base — but still a 61% improvement over the old 8.3%.

More interesting is what happens to 90-day retention. If van der Berg is right that healthy trial conversion should be followed by healthy retention, this is the real test. Manipulated conversions — people who convert because of engineered loss aversion rather than genuine product value — should churn fast.

Our 90-day retention for the new cohort is 78%. The old cohort's 90-day retention was 71%. The new users are not only converting at higher rates — they're staying longer. The integration-first onboarding isn't just creating psychological ownership. It's creating real product dependency. The users who connect their CRM, invite their team, and build workflows in the first week are getting genuine value from the product. The behavioral science didn't trick them into buying something they don't need. It helped them discover that they need it.

At least, that's what I tell myself. Van der Berg is honest about the ethical ambiguity of this kind of work, and I appreciate that. She writes that "if a significant percentage of users convert but cancel within the first three months, it is likely that loss aversion is doing too much of the work and genuine product-market fit is doing too little." By that standard, I think we're on the right side. But I'm also the person who designed the experiment, so I'm not exactly an unbiased observer.

Three things.

First: growth teams should be reading behavioral science. Not the pop-science summaries — the real research. Van der Berg's work is rigorous in a way that most marketing advice isn't. She cites sample sizes, acknowledges limitations, and distinguishes between lab findings and real-world evidence. That kind of rigor is rare, and it's useful.

Second: the experiment that almost didn't happen — because "thirty days is who we are" — turned out to be the highest-impact growth project we ran all quarter. Sacred cows are expensive. Question everything, especially the things nobody questions.

Third: the behavioral science didn't replace strategy. It informed it. We still needed to understand our users, our product, and our market. The research on loss aversion and hedonic adaptation gave us a hypothesis. The experiment confirmed it. But someone still had to design the onboarding flow, choose the integrations to prioritize, and make the case to the CEO. The science is the map. You still have to walk the territory.

I sent van der Berg an email when the results came in. I wanted to thank her, and I wanted to know if our numbers were consistent with what she'd seen in her research. She wrote back with three follow-up questions I hadn't thought to ask. I'm still working on the answers.